Weifeng Liu

Facial expression recognition based on Gabor features and sparse representation

In this paper, we present a facial expression recognition method based on Gabor feature and sparse representation. Sparse Representation based Classification (SRC) has been widely used in computer vision and pattern recognition. And Gabor filter banks can be used to approximately model the signal processing in visual primary cortex. We believe that the nature of the attractive performance of SRC and Gabor feature lies in that they both followed the natures of signal perception of retina and information processing of cortex in human vision. Therefore, we combined the Gabor feature and SRC for facial expression recognition. The comparison experiments of proposed Gabor+SRC algorithm and straightforward SRC application are conducted on JAFFE database. And the experimental results showed the attractive performance of the proposed Gabor+SRC method.

An effective eye states detection method based on projection

An effective multiple eye-state detection method based on projection is proposed. The structure of an eye sketch map is analyzed. Using integral projection method, the height and width of eye iris could be evaluated. And then the ratio between the height and width of the eye iris visible is selected as the criterion to determine the eye states. The experimental results showed the efficiency of eye states detection method proposed.

Facial expression recognition based on Hessian regularized support vector machine

Semi-supervised learning (SSL) has achieved attractive performance in many pattern recognition areas including image annotation, object recognition, face recognition and facial expression recognition. The state of the art SSL algorithm is Laplacian regularization (LR) which determined the underlying manifold using graph Laplacian. However, LR suffers from the lack of extrapolating power which will be towards the constant function for the data points beyond the boundary of domain. In contrast to LR, Hessian regularization (HR) can well steer the function varying smoothly along the manifold. In this paper, we present Hessian regularized support vector machine (SVM) for facial expression recognition (FER). We carefully conduct experiments on JAFFE dataset. The experimental results show that HR based SVM (HesSVM) outperforms SVM and LR base SVM (LapSVM).

Detection of static salient objects based on visual attention and edge features

Object detection algorithm based on the traditional saliency map often has problems of unable to locate and extract static salient objects precisely, besides, the position located does not coincide with the actual size of the object. In order to solve these problems, an improved algorithm of static salient objects detection is proposed in this paper, which combines visual attention mechanism with the edge information. First, coarse detection is realized according to an improved saliency map, then fine detection is implemented through edge detection by wavelet transform. Finally, mathematical logic operations are performed on the above two saliency maps, thus making the location and extraction of the static salient objects more accurately than traditional methods. Experimental results demonstrate the efficacy of the proposed method.

Subject-Independent Facial Expression Recognition with Biologically Inspired Features

Despite of much research for facial expression recognition, recognizing facial expressions across different persons is still a challenging computer vision task. However, facial expression analysis seems naturally for human visual system. Motivated by visual biology, this paper proposes an invariant feature extraction method for subject-independent facial expression recognition. In particular, we extract the biologically inspired facial features using extended visual cortex model-HMAX which consist of a template matching and a maximum pooling operation. We carefully organized the facial features and achieve subject-independent facial expression recognition using a sparse representation based classifier. The experiments on Yale database and JAFFE database demonstrate the significance of our proposed method for subject-independent facial expression recognition.

Moving object detection based on HFT and dynamic fusion

To solve the problem of detecting salient moving object in the video shot by static camera, a new spatio-temporal object detection algorithm is proposed in this paper. Firstly, Hypercomplex Fourier Transform (HFT) is used to the current video frame to achieve the static salient region; then, the moving salient region is detected by an improved three frames difference algorithm; finally, the static salient region and the moving salient region are combined with dynamic fusion strategy. Compared with the traditional techniques, the proposed method is in better correspondence with the response of the human visual system and more suitable for salient moving object detection. The experimental results demonstrate the effectiveness of the proposed object detection method.

Bio-inspired invariant visual feature representation based on K-SVD and SURF algorithms

In this paper, a bio-inspired invariant visual feature representation method is proposed. A set of Gabor filters with different parameters and global max operation are performed to improve the adaptability to scale and shift changes. In order to extract rotation-invariant features of images, the K-SVD and SURF algorithms are introduced into the traditional HMAX model. Prototypes (feature templates) are learned by the K-SVD algorithm, while the SURF descriptor of patches aims to enhance the rotation invariance. Experimental results on image classification demonstrate the superiority of the proposed feature representation method.

The comparison of different graph convolutional neural networks for image recognition

During the past decade, deep learning (DL) has been proven an effective way for image recognition. Various kinds of DL models such as, graph neural network (GNN), gated graph neural networks (GG-NNs) and graph convolutional neural networks (GCN) have been proposed. The latest GCN algorithm has achieved prominent performance for knowledge graph classification and molecular recognition. However, it only calculates the simple zero-one adjacency matrix, which is powerless to distinguish the self-node and other nodes. In this paper, we propose a novel negative one-zero-one graph to tackle this problem. Particularly, the negative one-zero-one graph employs zero to identify the node itself. To further explore the distance structure of inner-class samples, we develop the negative one-zero-one graph to a Euclidean distance version. Based on the ordinary graph, we also propose three kinds of hypergraph construction methods, such as hypergraph label method, hypergraph Euclidean distance method and the improvement method of hypergraph Euclidean distance. Additionally, we extend GCN to image recognition tasks in this paper. Extensive experiments are conducted on PASCAL VOC 2007 dataset to validate the proposed algorithms. The experiments results prove that the proposed methods is superior to the some methods for image recognition.

Dictionary learning based image enhancement for rarity detection

Image enhancement is an important image processing technique that processes images suitably for a specific application e.g. image editing. The conventional solutions of image enhancement are grouped into two categories which are spatial domain processing method and transform domain processing method such as contrast manipulation, histogram equalization, homomorphic filtering. This paper proposes a new image enhance method based on dictionary learning. Particularly, the proposed method adjusts the image by manipulating the rarity of dictionary atoms. Firstly, learn the dictionary through sparse coding algorithms on divided sub-image blocks. Secondly, compute the rarity of dictionary atoms on statistics of the corresponding sparse coefficients. Thirdly, adjust the rarity according to specific application and form a new dictionary. Finally, reconstruct the image using the updated dictionary and sparse coefficients. Compared with the traditional techniques, the proposed method enhances image based on the image content not on distribution of pixel grey value or frequency. The advantages of the proposed method lie in that it is in better correspondence with the response of the human visual system and more suitable for salient objects extraction. The experimental results demonstrate the effectiveness of the proposed image enhance method.

Object-of-interest extraction based on sparse coding

Usually, there are different kinds of targets in a complicated scene, but we human beings are only interested in some of them with salient or rare features, therefore how to detect and locate such objects of interest from a cluttered background is a key issue in computer vision research. In this paper, we propose an object-of-interest extraction method based on a rarity model derived from sparse coding. The rarity of an image is computed by analyzing the sparse coefficient matrix after dictionary learning and then used to extract the interested objects. Experimental results show that the proposed method has better performance than traditional methods.